fredsmith is mostly correct in his assertions, specifically that (with appropriate design) non-hardened (R grade) and non-space qualified (S grade) commercial parts can be used in satellites operating in LEO, at least if the software has good fault tolerance to hardware and you aren’t expecting a couple of decades of reliable life. It is also true that the handoff and aliasing schemes between one satellite and another, while complex to implement, are within existing experience for a reasonable degree of bandwidth, and in fact companies like Globalstar are already showing a robust capability which is being used by a few satellites for intraorbit communications (fleets of satellites communicating with ground) which is a significantly more difficult problem than calculating handoffs to a relatively stationary ground target.
The two points that bear consideration as remaining difficulties are this:
One, launch costs are still quite expensive, still pegging at >$5000/kg and going higher for smaller vehicles with dedicated vehicles. Yes, large launch vehicles can carry secondary payloads that pay a reduced cost, but often at the expense of not going to a desired orbit that is optimal for operations, which either requires considerable flexibility on the part of the spacecraft in terms of orbital parameters or onboard propulsion to phase into the desired orbit. Multiple payload deployments, such as are done with the Russian Rokot and US Minotaur IV vehicles can provide a dedicated ride for multiple satellites to some specific azimuth and altitude but still can’t just fly the payloads to optimal orbits for a global constellation. Upstarts like Space Exploration Technologies (SpaceX), RocketLab, and Firefly are promising massive reductions in space launch costs, largely through reusability, but this has yet to be demonstrated (SpaceX, by all public accounts, is still operating in the red and has to devote massive labor resources to every launch campaign which is not sustainable at the launch rates they need in order to achieve sufficient revenues) and historically the costs of reusability have been far greater than expected by orders of magnitude, and have not provided the expected reductions or increased availability compared to expendable vehicles. This isn’t to say that reduced launch costs won’t be realized–there is really no reason that the ground processing costs, which represent the bulk of reduce able labor can’t be minimized except for the costs of developing the necessary infrastructure to do so–but even today it is far from cheap to get a reliable ride to LEO, much less the hundreds that would be required for this effort.
The other is bandwidth. Certainly, at LEO the bandwidth concerns are somewhat mitigated by the smaller arc length that has to be uniquely occupied by a given satellite, and the fact that orbits at any azimuth can be achieved and sustained, giving overall greater coverage. But even if you fill the entire spectrum with overlapping signals, we’ll never be able to achieve the same bandwidth as a ground fiber optic cable network of comparable scale. Satellite broadband is a solution for areas where it is too costly to develop a ground infrastructure, but not a replacement for it, especially in dense urban areas.
Stranger